A Jellyfin server, plus audio and subtitle tracks

The headline is a Jellyfin server. Add your instance to a Jellyfin app and sign in with your configuration. Your catalogs become libraries, and the streams your addons find become versions to pick from. Continue Watching, Next Up and skip-intro buttons work as they do on any Jellyfin server, everyone in a household can have their own profile, and addons can act as trackers. A tracker records what you watch and brings back what you watched elsewhere.

For everyone, media info now keeps every audio and subtitle track a file was probed with. The formatter and stream expressions can ask for an English forced subtitle or a commentary track, not just a bare list of languages. The formatter can compare a stream against your own preferences, variants can insert anywhere in a list, and release groups are parsed more reliably, particularly on usenet. Busy public instances should also be more responsive at peak times.

For self-hosters, built-in Newznab and Torznab searches are faster, and each new feature comes with settings to limit it or turn it off.

A Jellyfin server

Your configuration can now be added to a Jellyfin app as a server. Nothing is imported, downloaded or stored. When you open or play something, AIOStreams asks your addons, exactly as it does for Stremio, and translates the answer into what a Jellyfin app expects.

Swiftfin, Findroid, Streamyfin, Jellyfin for Android TV, Jellyfin for Kodi and Infuse all work. The Jellyfin web interface does not, because the address is an API for apps, not a website.

Connecting

Open Save & Install, then the new Jellyfin card. Its Connect tab has everything an app asks for:

1. In the app, add a server and paste the Server address: your instance followed by /jellyfin, such as https://aiostreams.example.com/jellyfin.
2. Sign in with the Username shown there and your configuration password. The username is your configuration's UUID, or its alias if you gave it one through a profile.

On a TV, use Quick Connect instead. Pick it on the app's sign-in screen and type the six-digit code the app shows into the Connect tab. The tab also offers an address with a sign-in picker, which lists your configuration and its users on the app's sign-in screen with no password to type. That address contains your password, so only use it on your own household's devices.

Browsing

Each catalog becomes a library, in the order set under Addons → Catalogs. Searching in the app searches the catalogs that support search. A catalog of collections, such as a set of streaming services or film series, becomes a Collections library. Changes you save reach connected apps within about 30 seconds.

An instance limits how many libraries a configuration shows (20 by default) and how many titles each lists (250). The Connect tab shows your instance's numbers. Catalogs past the library limit do not appear at all, so move the ones you care about to the top.

Playing

Opening a movie or an episode searches your addons and lists what comes back as versions. They are named by your formatter and listed in your sort order, so pressing play without choosing plays the stream Stremio would put at the top.

The Playback tab controls this:

• Versions per item sets how many streams to offer, up to the instance's maximum (10 by default).
• Resolve streams when an item opens is on by default. Turn it off and nothing is searched until you press play, which is lighter on your addons.

Notices, errors and statistics your addons return are listed after the versions. Stremio shows these as links to open, which a Jellyfin app cannot do, so here you only see their text.

Streams are played directly, never transcoded, so your device has to handle the file as it is. If one version will not start, try another. A stream that needs custom request headers can only play through a proxy set up under Proxy, so without a proxy it is left out.

Skip intro and credits, on by default, gives apps intro, recap and credits markers, and the app shows a skip button. The markers come from community databases:

• IntroDB covers series, plus end credits for movies.
• AniSkip covers anime.
• PublicMetaDB covers movies as well as series. It needs an API key. If the instance has none, enter your own on the Playback tab.
• Anime Skip also covers anime, when the instance owner has set it up.

They are all asked at once. For each kind of marker, the first database on the list that has one supplies it. The Playback tab lists the databases your instance uses and lets you pick which markers to offer. Whether the app skips on its own is a setting in the app.

Watch history and trackers

AIOStreams keeps your progress, played marks and favourites itself. That is what fills Continue Watching and Next Up, and it is the same on every app you sign in to.

An addon that supports the new Watch State resource can act as a tracker. The Trackers tab lists the ones in your configuration and what each does:

• Records your plays. What you watch in a Jellyfin app is sent to the addon, so the watch-history service behind it stays up to date. Jellyfin apps report pause, stop and your position as you watch, so a tracker knows a finished episode from one abandoned a minute in. Stremio never tells an addon that.
• Imports your history. What the addon knows you watched elsewhere, including in Stremio, shows up in Continue Watching and Next Up.
• Keeps your watchlist in step with your favourites, when the addon has a watchlist.

To stop an addon syncing, untick Watch State in that addon's Resources option.

Household users

The Users tab adds profiles to your server, one per person. Everyone signs in with the configuration password, by typing their name or picking it from the sign-in picker.

• A user's history is either their own or shared with you.
• Variants can be applied per user, so one person gets different filters, or a different debrid account.
• With Choose trackers, each user with their own history syncs with their own tracker addons, such as a separate watch-history account per person.

Removing a user keeps their history. Add them back under the same name and it returns.

API keys

A tool that uses the Jellyfin API, such as a dashboard or a script, can connect with an API key instead of your password. Create one under API keys on the Connect tab. A key acts as the server's administrator: it sees every user of your configuration, including what each is watching right now. It stops working when you revoke it or change the configuration password.

For addon developers

When serving Jellyfin apps, AIOStreams reads more from addons than Stremio does:

• Extra metadata fields: credits, taglines, ratings, studios and library art.
• A collection block that turns a meta into a group of movies and shows.
• Subtitle titles, and forced and hearing-impaired flags.
• The watch_state resource that trackers implement.

Stremio ignores all of them, so an addon that uses them is still a normal Stremio addon. The addon protocol reference documents each one.

For everyone

Audio and subtitle tracks

Media info used to boil a file down to lists: its languages, its subtitle languages, its audio formats. It now keeps every track a probe found. A track has its language, title and codec, plus the format and channels for audio. It also carries the flags the file sets: default, forced, hearing impaired, commentary, dub, original language and audio description.

Tracks come from the same media info that already filled in accurate languages and subtitles: the probe data StremThru provides for debrid results from the built-in addons and service-wrapped addons. For other addons, an optional RemuxDB lookup can fill it in. A stream without media info has no tracks.

In the formatter, {stream.audioTracks} and {stream.subtitleTracks} are lists of tracks. Narrow them with ::where, take one field from each with ::pluck, or render each track with ::each:

{stream.subtitleTracks::where('hearingImpaired')::exists["CC"||""]}
{stream.subtitleTracks::where('forced')::pluck('lang')::languageEmoji::join(' ')}
{stream.audioTracks::each("{track.lang::languageCode} {track.tag} {track.channels}")::join(' | ')}

  ->  EN TrueHD 7.1 | EN DD 2.0 | FR DD+ 5.1

In stream expressions, audioTrack() and subtitleTrack() match a stream when a single track meets every condition. So an English track that is also forced counts, but an English track plus a separate forced track does not:

subtitleTrack(streams, 'lang=English', 'forced')
audioTrack(streams, 'lang=Japanese', 'original')
negate(subtitleTrack(streams, 'codec~pgs'), streams)

Many releases only mark SDH or commentary in a track's title, so title~sdh catches more than the hearingImpaired flag does. You can add and edit tracks on the sample stream in the formatter's preview panel. The formatter reference and the stream expression reference list every field.

The formatter

Your own preferences, as fields. The new {user.*} fields hold the lists you configured: languages, subtitles, resolutions, qualities, visual tags, audio tags, audio channels, encodes, stream types, release groups and keywords. Each one combines that list's preferred, required and included values.

Pass one of these fields in place of a value to ::where, ::in, or the new ::keep and ::remove, and a template can react to your configuration instead of repeating it. Change your preferred languages and the template follows.

{stream.languages::in({user.languages})["⭐"||""]}
{stream.audioTags::keep({user.audioTags})::join(', ')}
{stream.subtitleTracks::where('lang={user.subtitles}', '!forced')::exists["💬"||""]}

Also new:

• {stream.site} is the site a release names in its title, such as www.UIndex.org.
• ::trim strips whitespace, or any characters you give it, from both ends. With {stream.site} it tidies a filename: {stream.filename::replace('{stream.site}','')::trim(' -')}.
• Templates can be three times as long. The default limit rose from 5,000 characters to 15,000.

Variants

• insert puts a value anywhere in a list. add and prepend only reach the ends. Anchor it on a selector so it still lands in the right place after you reorder the list:

  insert after excludedStreamExpressions[expression*="Torbox Download Limit"] = {
    "expression": "/*No 4K*/ resolution(streams, '2160p')",
    "enabled": true
  }

• [*] also walks an object's values, not just a list's elements, so one instruction reaches every sort list you have defined. insert sortCriteria[*][1] = { "key": "resolution", "direction": "desc" } makes resolution the second criterion in each.

• Any number of variants can combine. The cap of four per request is gone. Instead, everything one request applies shares a single budget of instructions. The reference now spells out what applies, in what order when a variant chosen in the URL and variants activated by their conditions meet.

The reference has more on inserting at a position.

Release groups

An update to the release name parser means release groups are read more reliably. The parser also ignores the tags that reposters and indexers add after the group. These are everywhere on usenet: in Movie.2024.1080p.WEB-DL.x264-GROUP-AsRequested or ...-GROUP-Obfuscated, the release group is now GROUP, not AsRequested or Obfuscated. Release group filters, sorting and the formatter all see the real group.

Common suffixes are recognised out of the box, and instance owners can add more.

Steadier under load

Most of this release's performance work is aimed at busy public instances. Some jobs used to run from start to finish without letting anything else through: reading a large indexer response, ranking thousands of streams against a regex list, or going through a download with thousands of files. While one of them ran, every other request on the instance waited.

These jobs now run in short slices, so other requests keep moving while they run, and they use less CPU. A busy instance should see fewer slowdowns and smaller CPU spikes.

Smaller things

• PenguPlay is available as a preset: movies, series and live TV from public hosts, with provider and quality filters.
• SquareEyed is in the Newznab preset's indexer list, contributed by Spoorloos.
• Bitmagnet has a Search Mode option. It can search by ID where Bitmagnet supports it, always by title, or both. Title search was the only behaviour before, and it is still the default.
• Only Same-Release Failover, under Services → Failover, only fails over to other copies of the release you picked, never to a different release. Contributed by webstreamr.
• Also Check Result Age, in the Digital Release Filter under Filters → Miscellaneous, also blocks results uploaded before the title's release or air date, when the result's age is known. Contributed by webstreamr.
• There are four new stream expression filters: editions(), network(), container() and extension().
• RemuxDB Integration, under Filters → Miscellaneous, looks results up in RemuxDB's database of probed files. It fills in media info, tracks included, for addons that StremThru does not cover. It is off by default. Each lookup sends the title's IMDb ID, with the season and episode, from the instance. Contributed by webstreamr.
• Drafts can be checked before you restore them. The redesigned banner's Review button shows what restoring would change in the configuration you have loaded.

For self-hosters

Faster Newznab and Torznab searches

On an instance only you use, you are unlikely to notice the performance work, except in the built-in Newznab and Torznab searches:

• Responses are read about seven times faster. That takes roughly a tenth off a search that is not already cached.
• On Redis, their cache entries take about a twelfth of the space. Large cache values are now stored compressed. The biggest search responses, such as a full season of a long-running show, used to go over Max cached value size and were fetched again every time. Now they are cached.

New settings

• Jellyfin, under Settings → Jellyfin. Enable Jellyfin API (JELLYFIN_ENABLED) is on by default. Max libraries per configuration (JELLYFIN_MAX_LIBRARIES, 20), Max items per library (JELLYFIN_MAX_CATALOG_ITEMS, 250), Max versions per item (JELLYFIN_MAX_VERSIONS, 10) and Max users per configuration (JELLYFIN_MAX_PERSONAS, 20) set the limits. Resolve streams when an item is opened (JELLYFIN_RESOLVE_ON_OPEN) can force that behaviour on or off for everyone. Skip intro and credits (JELLYFIN_SEGMENTS_ENABLED, on) sends what is being played to the marker databases. Segment providers (JELLYFIN_SEGMENTS_PROVIDERS) sets which databases are asked and in what order. It defaults to IntroDB, AniSkip and PublicMetaDB. Anime Skip is only used with an Anime Skip client id (JELLYFIN_SEGMENTS_ANIME_SKIP_CLIENT_ID). A PublicMetaDB API key (JELLYFIN_SEGMENTS_PMDB_API_KEY) covers every configuration without its own. Leave it empty and PublicMetaDB is only used by configurations that bring a key. Logins, API calls and images each have their own rate limit: JELLYFIN_LOGIN_RATE_LIMIT_*, JELLYFIN_API_RATE_LIMIT_* and JELLYFIN_IMAGE_RATE_LIMIT_*.
• Watch State, under Settings → Watch State. Report playback to addons (WATCH_STATE_REPORT_ENABLED) and Read watch state from addons (WATCH_STATE_PULL_ENABLED) are both on. Max addons exchanged with (WATCH_STATE_MAX_SINKS, 3) limits how many trackers a configuration can use. A tracker addon on a private address also needs Allow exchanging with private addresses (WATCH_STATE_ALLOW_PRIVATE_URLS, off). That lets anyone who can save a configuration make your server send requests into your network, so only turn it on for users you trust.
• Repost suffixes, under Settings → Addon Resources (REPOST_SUFFIXES), is the list of tags the parser ignores after a release group. End an entry with * to match anything that starts with it.
• Enable release blocklist, under Settings → Blocklists (RELEASE_BLOCKLIST_ENABLED, on), switches off the whole release blocklist in one go. While it is off, nothing is filtered, no new verdicts are recorded and subscribed lists stop syncing. Existing entries are kept. Contributed by webstreamr.
• Max linked accounts per configuration, under Settings → General (LINKED_ACCOUNTS_MAX_PER_USER, 10).
• RemuxDB, under Settings → RemuxDB (REMUXDB_ENABLED, on), makes the RemuxDB lookup available to users. Lookups go out from your server. Turning this off hides the option and stops them.
• DATABASE_POOL_SIZE (10) sets how many PostgreSQL connections the instance opens. It is a startup setting with no dashboard equivalent. Requests, background tasks and watch-state imports all share this pool, so an instance serving many Jellyfin clients wants a bigger one.

Upgrading

Nothing in this release changes install URLs or needs a manual migration. A few things are worth checking.

Network names. The parser now uses the names the services use for themselves. Amazon is now Prime Video, Disney is Disney+, HBO is HBO Max and Paramount is Paramount+. Update any formatter template that compares {stream.network} against the old names.

Variant limits. MAX_VARIANT_INSTRUCTIONS, MAX_ACTIVE_VARIANTS and MAX_VARIANT_DEPTH are gone. No per-script cap could bound what nested use variant instructions expand to, so one budget now covers everything a request applies: Settings → User Limits → Max variant instructions per request (MAX_VARIANT_TOTAL_INSTRUCTIONS, 5000). A request that runs out stops there and says so. The old variables are ignored.

Private sync URLs. Regex and stream expression sync URLs on private addresses are now refused unless Settings → User Limits → Allow private regex/SEL sync URLs (SYNC_ALLOW_PRIVATE_URLS) is on. Whitelisted URLs you configured yourself are not affected.

Newznab and Torznab caps. Built-in Newznab and Torznab searches take at most 1,000 results and 10 MB per response. Change these under Settings → Built-ins with Newznab/Torznab max results per request (BUILTIN_NAB_MAX_RESULTS) and Newznab/Torznab max response size (BUILTIN_NAB_MAX_RESPONSE_BYTES). An indexer that returns more is cut off at the cap with a warning, and the search still succeeds.

Several replicas on one Redis. Large cache values are now stored compressed. The new version reads old values fine, but an older version logs a parse error every time it reads a compressed value, until that value expires. Upgrade every replica together.

A database from another build. If a database's migrations were applied by a different build of AIOStreams, such as a fork or a development branch, startup now refuses it and names the migrations that conflict. Before, the instance ran on a schema it did not expect. To fix it, point DATABASE_URI at a fresh database or go back to the build that created it. Upgrading from a release is not affected.

What is on by default. The Jellyfin server at /jellyfin, with skip markers and watch state in both directions. RemuxDB is also available by default, but each user has to turn it on for themselves. New settings says where to switch each one off.